The estimation and acquisition of traffic parameter information is the key to solving urban management and control problems. However, traditional methods are difficult to obtain traffic parameters efficiently and accurately in complex traffic scenarios. The rapid development of information technology has brought new directions to the solution of traffic management and control problems. This paper proposed a novel video-based traffic parameter extraction system which consists of two parts: analysis of traffic parameters in a traffic video and trajectory processing. In the first part, we used advanced techniques such as deep learning, calibration method and image processing to obtain the key information such as vehicle trajectories of the traffic video. In the second part, all information of the first part was processed uniformly and generated traffic parameters such as traffic flow, vehicle type, vehicle composition of different vehicle types, and speed of vehicles passing through a scene in a traffic video. The experimental results show that the accuracy of the detailed traffic flow information obtained by the proposed system can reach more than 90%, the accuracy of vehicle composition of different vehicle types can be achieved more than 98%, and the vehicle speed accuracy can reach more than 85%. High-precision and abundant traffic parameters can provide important data support for traffic management and control, which illustrate the importance and significance of the proposed system.

CICTP